INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Adenovirus serotype 5 (Ad5) is the most commonly used vector for gene therapy because it demonstrates an outstanding efficacy of gene transfer in vivo; it infects both proliferating and highly differentiated cells. Ad5 grows to high titer, and large (up to 6.5-kb) foreign DNA fragments can be incorporated into the Ad genome serving as a transgene. However, Ad as a gene therapy vector also has disadvantages, including the broad distribution of the Ad primary receptorthe coxsackievirus-Ad receptor (CAR)which precludes specific gene delivery. In addition, many malignant cell types lack the CAR and are therefore not permissive for gene therapy with nontargeted Ad vectors (for a review, see reference 18). Ad retargeting, that is, redirecting the viral infection to certain cells specifically, is therefore one of the major areas being addressed by many investigators in the field (3).

A number of strategies have been developed to achieve targeted gene delivery with Ad vectors. Two general approaches are currently used to modify the natural tropism of Ad. One approach in Ad targeting is to use bispecific molecular "bridges" (chemical or genetic fusion conjugates), one end of which specifically binds a virus capsid protein whereas the other end binds to a cellular marker (5, 6, 8, 9, 15, 21). The other approach is genetic modification of the virus particle itself, thereby incorporating specific targeting ligands directly into Ad capsid proteins, which in turn permits Ad to acquire expanded tropism. Since the Ad fiber protein, and its carboxy-terminal knob domain in particular, plays the major role in virus-cell interaction (12), this protein is a reasonable site for specific ligand incorporation. Two distinct locales within the Ad knob domain have been employed to modify viral tropism: the carboxy terminus (13, 22) and the HI loop of the fiber knob (4, 10, 23). A critical consideration in generation of Ads with modified knobs is the need for the knob fiber to retain its natural ability to form trimers. Therefore, knob-ligand structural compatibility is one of the key issues to be addressed while creating genetically modified Ad vectors. In addition, promising candidate ligands very often lose their fidelity as targeting moieties once they are introduced into the Ad virion. Thus, the two issues of ligand structural compatibility and "in-context" fidelity are critical.

A promising way to identify potential targeting moieties is to exploit a high-throughput approach, such as screening of phage-displayed ligand libraries. However, considering the above-mentioned issues, for development of new targeted Ad vectors it would be desirable to improve such an approach by combining the advantage of high-throughput phage library screening with in-context ligand functional and structural suitability. This could be achieved by screening ligand libraries incorporated directly into the Ad5 knob domain displayed on the surfaces of bacteriophages. However, the conventional filamentous-phage display allows only amino-terminal insertions into the product of gene III (20), whereas the Ad knob is the C-terminal portion of the fiber. To circumvent this obstacle, we decided to employ a phage display system, pJuFo (2), which was originally designed to display C-terminal protein fragments. This system explores a strong association of the Jun and Fos leucine zipper domains. The vector features simultaneous production of two recombinant proteins: phage protein pIII fused with the Jun polypeptide and the cDNA product fused with Fos. Both proteins are transported into the periplasm, where the Jun-Fos association occurs followed by stabilization of the heterodimer by two disulfide bonds (Fig. 1A). The recombinant pIII carrying a covalently attached cDNA product is then incorporated into the phage body, and thus, the product becomes phage displayed (Fig. 1B). Wild-type pIII is provided in trans by a helper phage and is necessary for phage infectivity. Theoretically, phagemid-based systems allow up to five copies of recombinant protein to be displayed on one phage particle. Our hypothesis was that neighboring knobs would form trimers on the phage surface and such phage-displayed trimeric knobs would therefore represent a tool for studying potential ligand structural compatibility and fidelity (Fig. 1C). In the present proof-of-principle study, we describe phage display of functional Ad5 knobs in the pJuFo system and address the issues of ligand-knob fidelity and ligand-knob compatibility.

View larger version (33K): [in this window] [in a new window]   FIG. 1.   Display of the Ad5 knob on the surface of filamentous bacteriophage. For a detailed explanation, see the text.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Phage display. The pJuFo phagemid vector was obtained from Reto Crameri (Swiss Institute of Allergy and Asthma Research, Davos, Switzerland). The bacterial host for cloning and phage propagation was Escherichia coli XL1 Blue (Stratagene, La Jolla, Calif.). The helper phage was VCSM13 (Stratagene).

Cloning and phage display of Ad fiber knobs. General microbiological and molecular biological techniques and recipes for bacterial growth media are described in reference (19). The wild-type Ad5 fiber knob gene fragment coding for amino acid positions 386 to 581 of the fiber (according to reference 1), including the last fiber shaft repeat and the entire Ad5 knob, was PCR amplified from the plasmid pNEB.PK3.6 (11) using the primers Xba_knob_for (5'-ATA TCT AGA ACA GGT GCC ATT ACA GTA GGA A-3') and knob_Kpn_rev (5'-AAC GGT ACC TTA TTC TTG GGC AAT GTA TGA A-3'). The XbaI and KpnI restriction sites (underlined) were introduced into the amplification product. The PCR product was digested with XbaI/KpnI and cloned into the pJuFo vector also cut with XbaI/KpnI. This resulted in phagemid pJuFo.w.t.kn.

View larger version (42K): [in this window] [in a new window]   FIG. 2.   Construction of pJuFo.kn.2Esp vector for unidirectional cloning within the Ad5 knob HI loop. (A). The vector was designed to allow peptide insertions without disruption of the native Ad5 knob amino acid composition. Twelve base pairs coding for a Gly(543)-Asp-Thr-Thr(546) oligopeptide were deleted, and two adjacent Esp3I restriction sites were introduced into the opposite DNA strands of the knob gene. Two portions of the knob gene were PCR amplified from pJuFo.w.t.kn and connected by blunt-end ligation. The reverse primer for the left fragment was designed to substitute the triplet GAA coding for Glu(541) for GAG and the triplet ACA coding for Thr(542) for ACG, thus introducing an Esp3I restriction site into the complementary DNA chain. The forward primer for the right fragment was designed to introduce another Esp3I site into the plus-DNA chain and to substitute the triplet CCA coding for Pro(547) for CCT, which allowed the formation of an AvrII-compatible cohesive end after the phagemid was cleaved with Esp3I. (B) Cloning of the c-Myc epitope into the HI loop of the Ad5 knob displayed on phage. The stuffer sequence is removed by Esp3I digestion and replaced with an oligonucleotide duplex.

Immunological methods. Phage yield, the presence of trimeric knobs, and the solvent accessibility of knob-displayed ligands were assayed by dot blot analysis of phage displaying either the wild-type (wt) knob or ligand-containing knobs. The parent pJuFo phage displaying no knob was used as a negative control. One-microliter aliquots of the phage were applied to nitrocellulose membranes in parallel in twofold serial dilutions, starting from 10⁸ CFU/dot. The membranes were dried in air, blocked with TBST-casein (Tris-buffered saline [10 mM Tris-HCl, pH 7.4, 150 mM NaCl] plus Tween 20 to 0.05% plus casein to 0.5%), and incubated with antibodies. A monoclonal antibody (MAb) against M13 phage coat protein pVIII conjugated with horseradish peroxidase (HRP) (Amersham Pharmacia Biotech, Piscataway, N.J.), MAbs against polyhistidine and the c-Myc epitope (clones HIS-1 and 9E10, respectively, Sigma, St. Louis, Mo.), and the MAb 1D6.14, developed in our laboratory (6) and recognizing only the trimeric form of the knob, were used. For unconjugated antibodies, the membranes were additionally treated with goat anti-mouse immunoglobulin G (IgG)-HRP conjugate (DAKO, Carpinteria, Calif.). The color reaction was developed by incubation of the membranes with Sigma Fast diaminobenzidine (Sigma).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Cloning and phage display of Ad fiber knobs. The pJuFo system was used to display the unmodified wt Ad5 fiber knob (amino acids 386 through 581) and knobs incorporating heterologous ligands on the surfaces of filamentous bacteriophages. The modifications included incorporation of (i) six consecutive histidine residues at the knob C terminus [pJuFo.kn.6H(C)], (ii) the c-Myc epitope (EQKLISEEDL) at the knob C-terminus [pJuFo.kn.myc(C)], and (iii) the c-Myc epitope within the knob HI loop [pJuFo.kn.myc(HI)]. To clone the c-Myc epitope sequence into the HI loop without disruption of the native knob amino acid composition, an intermediate phagemid, pJuFo.kn.2Esp (Fig. 2), was designed and constructed. The correct gene assembly was confirmed by DNA sequencing. The recombinant phages pJuFo.w.t.kn, pJuFo.kn.6H(C), pJuFo.kn.myc(C), and pJuFo.kn.myc(HI) were rescued from E. coli XL1 Blue bacteria transformed with the phagemids by coinfection with a helper phage. Polyethylene glycol-double-purified phage were examined in a number of assays. We wished to demonstrate that the recombinant bacteriophages displayed recombinant knobs on their surfaces, that the displayed knobs adopted a native trimeric conformation in the context of the phage particle, that the displayed knobs are functional (they retain the ability to recognize and bind the primary Ad5 receptorthe CAR), and the ligands displayed in the context of such functional knobs are exposed and thus are able to direct knob-displaying phage to a specific target.

Examination of phage-displayed knobs. The dot modification of immunoblotting was employed to analyze the presence of the trimeric knobs at the phage surface and the ligands' accessibility. The rationale for using this method was based on the ability of nitrocellulose to adsorb proteins with high efficiency, independent of protein sorption properties, which could influence direct phage fixation on plastic. Phages displaying the wt knob, knob.6H(C), knob.myc(C), and knob.myc(H) were tested. Parent pJuFo phage displaying no knob were used as a negative control. Aliquots of twofold serial phage dilutions starting from 2 × 10⁸ CFU/dot were applied to nitrocellulose membranes, followed by treatment with anti-bacteriophage, anti-trimeric knob, and anti-ligand (six-His and c-Myc) antibodies. The results, presented in Fig. 3, demonstrated that the phages were used in approximately equal amounts (treatment with anti-M13 MAb [Fig. 3A]), that all the phage-displayed knobs interacted with MAb 1D6.14, recognizing only the trimeric knobs (Fig. 3B), and that both the c-Myc epitope (treatment with anti-c-Myc MAb [Fig. 3C]) and six-His (treatment with anti-polyhistidine MAb [Fig. 3D]) phage displayed in the context of the Ad knob were antibody accessible. It should be noted here that somewhat abrupt decrease in dot intensity in the last visible phage dilutions is routinely observed in our other dot assays and seems to be a comon feature of this kind of analysis.

View larger version (68K): [in this window] [in a new window]   FIG. 3.   Dot blot analysis of phage-displayed Ad5 knobs. One-microliter aliquots of the phage were applied to nitrocellulose membranes in parallel in twofold serial dilutions starting from 2 × 108 CFU/dot. The membranes were dried in air, blocked with TBST-casein, and incubated with antibodies. When necessary, the membranes were additionally treated with goat anti-mouse IgG-HRP conjugate. The color reaction was developed by incubation of the membranes with Sigma Fast diaminobenzidine. (A) Treatment with mouse anti-M13 MAb-HRP conjugate; (B) treatment with mouse anti-Ad5 knob MAb 1D6.14; (C) treatment with mouse anti-c-Myc MAb 9E10; (D) treatment with mouse anti-polyhistidine MAb HIS-1. Lanes: 1, pJuFo phage displaying no knob; 2, pJuFo phage displaying wt knobs; 3, pJuFo phage displaying knobs with six His at the C termini; 4, pJuFo phage displaying knobs with c-Myc at the C termini; 5, pJuFo phage displaying knobs with c-Myc within the HI loop.

View larger version (36K): [in this window] [in a new window]   FIG. 4.   Immunoblot of phages displaying Ad5 knob. Each phage sample was dissociated in sodium dodecyl sulfate sample buffer and divided into two tubes. One tube from each pair was heated at 96°C for 5 min, and the other was left unheated. Phage samples (1012/lane) were separated on 4 to 15% gradient PAAG followed by electrotransfer onto PVDF membranes. After being blocked with TBST-casein, the membranes were treated with antibodies followed by treatment with goat antimouse IgG-HRP conjugate. The color reaction was developed by incubation of the membranes with Sigma Fast diaminobenzidine. (A) Treatment with anti-Ad5 knob MAb 1D6.14; (B) treatment with anti-polyhistidine MAb HIS-1; (C) treatment with anti-c-Myc MAb 9E10. Lanes: N, pJuFo phage displaying no knob; W, pJuFo phage displaying wt knobs 6H, pJuFo phage displaying knobs with six His at the C termini; MC, pJuFo phage displaying knobs with c-Myc at the C termini; MHI, pJuFo phage displaying knobs with c-Myc within the HI loop; b, boiled sample; ub, unboiled sample.

View larger version (26K): [in this window] [in a new window]   FIG. 5.   Phage displaying Ad5 knob interaction with CAR in ELISA. Recombinant shCAR was adsorbed overnight at 4°C in wells of a 96-well MaxiSorp immunoplate at 200 ng/well in 0.1 M bicarbonate buffer, pH 9.1. After the plates were washed with TBST and blocked with TBST-casein, phage were applied to the wells in serial dilutions in TBST-casein, starting from 2 × 1011/well, and incubated with shCAR for 1 h at RT. Phage-shCAR interaction was revealed by treating the wells with the set of antibodies described in the legend to Fig. 3. The color reaction was developed by incubation with orthophenilenediamine. (A) Treatment with mouse anti-M13 MAb-HRP conjugate; (B) treatment with mouse anti-Ad5 knob MAb 1D6.14; (C) treatment with mouse anti-c-Myc MAb 9E10; (D) treatment with mouse anti-polyhistidine MAb HIS-1. All binding reactions were done in triplicate, and the mean values of the three measurements were taken as end points. term., terminus.

Phage displaying six-His knob bind to Ni-NTA. We have examined the potential ability of the six-histidine tag present at the C terminus of the phage-displayed knob in the construct pJuFo.kn.6H(C) to interact with the Ni-NTA affinity matrix. The resin was incubated with 10⁹ phage CFU and washed, and the bound phage was eluted by pH shift. As shown in Fig. 6, in contrast to the negative control and to the phage displaying the wt knob, the six histidines have mediated specific phage binding to the Ni-NTA. Three orders of magnitude binding preference was observed.

View larger version (21K): [in this window] [in a new window]   FIG. 6.   Phage displaying six-His knob bind to Ni-NTA. Phage (109 CFU) were mixed with 50 µl of Ni-NTA resin in 3 ml of PBS containing 10 mM imidazole, incubated for 1 h at RT with end-over-end rocking, washed five times with the incubation buffer, and eluted with 200 µl of 0.1 M Tris-glycine, pH 2.5. The eluate was immediately neutralized by the addition of 200 µl of 1 M Tris-HCl, pH 7.4, and used to infect E. coli followed by plating the infected cells on ampicillin-containing agar plates. The amount of bound phage was calculated by colony counting. The binding reactions were done in triplicate, and the mean value of the three colony countings was used.

Phage-displayed-knob interaction with U118/HisAR cells. U118 and U118/HisAR (anti-six-His scFv) cells were treated with pJuFo phage displaying sixHis at the C terminus of the knob, followed by development with anti-bacteriophage MAb. The staining results are presented in Fig. 7. It can be clearly seen that bacteriophage displaying the Ad5 knob containing the six-histidine sequence at the C terminus specifically interacted with the cells expressing an anti-polyhistidine scFv on their surfaces (Fig. 7B). Staining the U118/HisAR cells with an antibody recognizing an HA epitope present within the AR molecule revealed a similar pattern (Fig. 7D). Phage displaying the wt knob did not react with either U188 or U118/HisAR cells (not shown).

View larger version (143K): [in this window] [in a new window]   FIG. 7.   Phage-displayed knob interaction with U118/HisAR cells. CAR-deficient U118/HisAR cells and their parent U118 cells were grown to near confluence in the wells of a 24-well tissue culture plate; the growth medium was removed, and the cells were incubated for 1 h at RT with 1011 phage displaying six-His knobs in 500 µl of fresh growth medium. As a control, the cells were also treated with MAb HA-7 against the HA epitope. The cells were gently washed three times with the growth medium and fixed with 10% formaldehyde in PBS for 10 min at RT. Phage interaction with the cells was revealed by treatment with an anti-M13 MAb. The color reaction was developed with the DAKO EnVision system, which gives bright red staining. The nuclei were counterstained with hematoxylin aqueous formula. (A) Staining of U118 cells with pJuFo.kn.6H(C); (B) staining of U118/HisAR cells with pJuFo.kn.6H(C); (C) staining of U118 cells with MAb HA-7; (D) staining of U118/HisAR cells with MAb HA-7.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

If Ad vectors are to achieve their full potential as gene delivery agents for the treatment of human disease, they must be capable of specific target cell transduction. Both conjugate-based and genetic approaches have been tried to improve the specificity of Ad infectivity, with each achieving some measure of success (3). However, at this time there is still a critical lack of a vector which has the degree of targeting fidelity required for many clinical applications. Development of genetically modified vectors is attractive because it allows gene therapy with a single integral agent, a modified virus, thus eliminating the potential complication of the two-component conjugate approach. The development of such a vector is, however, limited by the structural constraints of modifying Ad vectors. In the present report, we present a novel approach for the definition of targeting ligands which assures a priori structural compatibility and targeting fidelity. This approach thus has the potential to substantially improve the development of targeted Ad vectors.

Phage display is a powerful tool to identify ligands that specifically recognize and bind a target of interest (16). By screening vast libraries of ligands, a specific ligand that binds to a majority of targets can be identified. Phage display has already been shown to allow the definition of ligands which have demonstrated subsequent utility for Ad-mediated gene delivery (14, 23). We attempted to demonstrate that the power of the high-throughput phage display approach could be combined with selecting the target-specific ligands in the context of an Ad5 knob. In this pilot study, we displayed the Ad5 knob on the surfaces of filamentous bacteriophages and demonstrated that ligands displayed in the context of phage-displayed knobs are able to recognize their targets.

We have demonstrated that the phage-displayed knobs were trimeric, as shown by the positive interaction of the knob-displaying phages with a MAb recognizing only the trimeric knob conformation (Fig. 3 to 5). In addition, the trimer formation was demonstrated by direct phage immunoblotting (Fig. 4). The knobs were shown to remain functionally active: they all retained the ability to recognize and bind the Ad5 receptor CAR (Fig. 5). It is important to note that the observed knob-shCAR interaction was developed with treatment by an anti-bacteriophage antibody. These results strongly suggest that the knobs are physically connected to the phage body and minimizes the likelihood that such an interaction may be mediated by loose knobs present in the prokaryotic phage preparations. The current model of Ad-CAR interaction is that the fiber can only bind the CAR as a trimer (17). In this study, the positive knob phage-shCAR binding results added to our confidence that the phage-displayed knobs are trimeric.

An important issue in our investigation was whether a ligand situated in the context of the phage-displayed Ad5 knob retains its target-binding capacity. As was shown in our experiments with selective binding of six-His-tagged phage-displayed knobs to Ni-NTA, we observed at least 3.5 orders of magnitude preferential binding in comparison to phage displaying the wt knob and to the negative control phage (Fig. 6).

Finally, we have demonstrated the ability of a phage knob-displayed ligand to interact with its specific receptor presented on living cells. The immunocytostaining results presented in Fig. 7 clearly show that the phage-cell binding pattern was similar to that of a MAb to the HA epitope present on the AR molecule.

Taken together, the data support the concept that phage display of the Ad5 knob provides a useful tool for investigation of ligand-knob fidelity and structural compatibility. Ligand candidates can be tested in this phage display system prior to the generation of an actual modified Ad. Moreover, screening libraries of random peptides using the phage display system described here will allow the selection of ligands that are knob (i.e., Ad) compatible and that retain fidelity. One important issue is whether the ligands identified using our knob phage display system will retain fidelity in the context of modified Ad. The ligand-screening procedure is twofold: screening for specificity followed by screening for compatibility (trimer formation). Since the fiber-trimerizing forces on the virion are much stronger than those on our artificial system (there are several shaft repeats stabilizing the fiber trimer compared with just one shaft repeat in our knob-pJuFo system), even weak trimers on phage should reflect much more stable trimers formed on the virion.

In the present report, we have established the feasibility of displaying peptide ligands in the context of the Ad5 knob on the surfaces of bacteriophages. This will enable the generation of knob peptide libraries suitable for screening against both purified proteins and cells. For maximal utility, however, especially in the context of library screening against cells (which may express various amounts of CAR), the knob displayed on the phage will need to be mutated to ablate its natural recognition of CAR. This can easily be achieved by simple point mutation (17) without disturbing the region into which the peptide libraries are inserted. In this way, the applicability of the approach may be greatly improved, since ligand selection will not be compromised by CAR recognition.